METHOD FOR EXTRACTING DISTINCTIVE FEATURE USING COLOR FACE IMAGE DATA TENSOR FACTORIZATION AND SLICING

摘要

A feature point extracting method using slice selection and color face data tensor analysis is provided to select a slice strongest to a change of illumination from a 3D feature point tensor, thereby stably extracting a feature point strong to the change of the illumination. Processes of sequentially optimizing projection matrices A, B, and C and a feature point tensor G in turn are repeatedly performed until calculated values are collected(310-316). A 4D cube tensor is configured from color face image data. The 4D cube tensor is analyzed to project a matrix of a large-sized 3D cube into a matrix of a small-sized 3D cube. Among slices of an optimized feature point tensor of the projected and calculated 3D cube, the third slice strong to a change of illumination is selected as a feature point.